|
Abstract:
As part of the Astrometry.net project, we are developing a solver for the
"blind astrometry" problem: given an astronomical image, where on the sky
did it come from? We are using a feature-matching approach: during the
preprocessing stage, we build an index of the sky starting from a
catalogue of known stars. Given a new test image, we extract features
from it and match them to features in the index. We treat stars as having
basically no distinctive features other than their positions, so we build
up geometric features which describe the relative positions of small sets
of stars.
I'll first show some rather impressive results from our system. We want
to build on this success in two ways: first, using this basic engine we
want to build some useful, compelling and cool tools for amateur and
professional astronomers. Second, we want to generalize this approach and
try applying it in other domains, by thinking about the properties of
problems or domains where purely geometric features could be useful.
This is at a very preliminary stage, so we're hoping to solicit some ideas
from the audience!